Kneron to Bring AI Enabling Solutions & AI-Embedded PC to COMPUTEX 2024

Connecting AI will be the main theme of COMPUTEX 2024 and as companies increasingly integrate AI into their products or solutions, AI enabling technologies will become more important for multiple types of companies. AI has seen a recent boom with the release of ChatGPT and other generative AI (GenAI) models. However, this rapid growth in interest also comes with concerns such as the required power and data for training the AI models as well as the potential for “AI hallucinations,” which can potentially give users false results.

During a visit to the Kneron Taipei office, the COMPUTEX team had the opportunity to chat with Dr. Albert Liu, the founder and CEO of Kneron. Kneron is a San Diego-based full stack AI company known for pioneering neural processing units (NPUs). With a focus on developing integrated edge AI hardware and software solutions for vehicle, security, and broader AIoT use cases, Kneron aims to empower the proliferation of accessible, low-latency, and secure AI applications by creating networks of independently intelligent devices as enabled through full stack Kneron solutions.

During the interview, Dr. Liu shared some of his experience in addressing the current issues with AI. “There are ways to reduce the cost of AI, such as by repurposing existing AI models. If they perform similar tasks and analyze similar data, they can be used for more than one specific purpose. For example, a model for smart car applications that has to consistently analyze road conditions and vehicles might also be usable in smart surveillance cameras.”

In regard to concerns on high hardware specs and power requirement, Dr. Liu said, “Most AI models today run online in servers equipped with high-end GPUs. While this is a viable option, it can be too expensive for startups or smaller companies who might have limited budgets. For example, the current “holy grail” for most AI developers or AI-integrated companies is the NVIDIA H100, but it is not only rare, it is also expensive.

“To train their AI, especially their GPT-based solutions, many companies today upload their training data to publicly available vendors such as OpenAI to help reduce their costs. Which means the company needs to properly vet and filter the data to make sure they do not break any confidentiality agreements or risk AI hallucinations if their training data was somehow affected by another dataset. In addition, because the data is used to train a public AI model, it might be leaked, possibly leading to legal concerns.”

This dilemma of confidentiality vs cost has encouraged Kneron to release their alternative: the KNEO-300, specially designed for enterprise GPT applications. The KNEO-300 is an NPU based edge Al server with high performance and low energy consumption, which can be applied in various enterprise GPT scenarios. It can be used for both GPT model training services and offline AI servers for faster deployment.

“Because the KNEO-300 is geared for offline deployment, developers can make sure their training data is secure and not contaminated by irrelevant data.” Dr. Liu stated. The offline feature also means that sensitive data such as legal information, medical history, trade secrets, etc. will remain safe under the control of the users. This means by using KNEO-300, users can develop GPT applications for various industries that require discretion and confidentiality such as legal, medical, accounting, startups, etc.

Established companies are not the only ones interested in GPT solutions as startups are also exploring the possibilities of developing their own specific GPT applications. To achieve this, startups might want to consider utilizing Kneron chips in EDGE GPT. With their focus on providing customers with computational infrastructure and creating user-friendly toolchains and ecosystems, users should be able to easily utilize their chips. Currently, Kneron has customers and partners all over the world, including Qualcomm, Sony, Toyota, Hanwha, Panasonic, and Foxconn.

AI in Every Device

Recent trends in the automotive industry have shown that AI-enabled vehicles will be a reality in the near future. Kneron’s AI chips are now already deployed in vehicles developed by multiple manufacturers, including being integrated into Toyota's onboard devices. Kneron has also collaborated with a number of major international automotive companies, including the recent DMS demonstration by oToBrite at the Taipei International Automobile Electronics Show.

While the applications of AI in vehicles will be varied and significant, one of the key technologies will be advanced driver assistance systems (ADAS). As autonomous driving and large language models continue to advance, the future will trend towards integrating AI in more vehicles. Kneron's high-performance AI SoC for edge computing offers quick response, low latency, low power consumption, and excellent ISP performance. It can also effectively adapt to complex lighting conditions during driving and support intelligent cabin interaction by processing computations locally in the vehicle. Its powerful algorithms can also identify people, vehicles, roads, and objects during driving, providing robust AI support for vehicles.

The AI PC Era is Coming

The demand for AI today is no longer limited on the organization level because many individual users are considering AI-empowered hardware for their daily use. While this might be possible during online operations, it might not be feasible when connections are poor or unavailable.

During the interview, Dr. Liu also mentioned one of the main issues with edge AI. “Most edge AI models today run on GPUs, which means the system has 3 vital requirements: large space, high energy, and plenty of cooling. As the pioneers and the trademark holder for NPUs, I think it would make sense for Kneron to put NPUs inside devices like PCs, laptops, smartphones, or even cameras because it is not limited by those 3 points.” Dr. Liu said.

Kneron’s latest generation AI chip already supports the Transformer architecture. The Kneron KLL830 that will be released this year provides more comprehensive support for Transformers and large language models. Lightweight large language models (LLM) can even be directly deployed on the KLL830. With Kneron’s unique multi-chip cascading technology integrates multiple small chips to provide computational power for 8B and 13B models and above, while maintaining cost advantages. In the future, Kneron will continue to develop their KL1140 chip, which is expected to have high computational power. This will enable it to cover large language model applications in edge servers and cloud environments.

“AI PC at the moment is in an interesting turning point because everyone knows it is coming, but it is not yet fully defined. This means whoever makes the most significant progress in this industry will be able to define it.” Dr. Liu stated.

Join Kneron in COMPUTEX 2024 and COMPUTEX Special Events

To share the ongoing trends of AI PC and GenAI, Kneron founder and CEO, Dr. Albert Liu will join the AI PC Industry Exploration Forum on May 16 with speakers from various leading companies in the AI ecosystem. In his keynote speech, Dr. Liu will discuss how NPU makes personalized GPT possible. Read the pre-event press release here:
>>> https://show.computex.biz/NewsReleaseDetail.aspx?index=42523&category=68

Kneron will also join COMPUTEX 2024 in booth N1223, showcasing various leading client products based on their latest technology and advanced solutions at booth N1223 in the AI Computing and System Solutions Area. Kneron’s highlight will include: Offline and Private GPT Solutions, AI PC Solutions, AI Accelerator Card (GPU+NPU), Intelligent Driving for Automotive, AIoT, and more.

Dr. Liu will be one of the speakers in the Exploring the Future of AI Forum in InnoVEX 2024. The forum will take place in the InnoVEX Center Stage on June 5, 15:30 – 17:30. Save your spot here:
>>> http://innovex.computex.biz/show/forum.aspx?id=43

■《About COMPUTEX TAIPEI》

COMPUTEX TAIPEI was founded and named by the then Chairman of Taipei Computer Association (TCA), Stan Shih. In 1985, TCA invited TAITRA to be a co-organizer of COMPUTEX TAIPEI. In 2016, the startup focused event, InnoVEX was introduced.

●COMPUTEX TAIPEI 2024 Overview●

This year’s special highlights:
Responding to the rapid development of AI technologies and applications including GenAI and LLM as well as the continued increase in the global digital transformation demands; this year’s COMPUTEX 2024 will focus on “Connecting AI”. This year, COMPUTEX will be joined by major semiconductor and ICT manufacturers including: Acer, ADATA, AMD, ASPEED, ASRock, ASUSTeK, ATEN, BenQ, Clientron, Cooler Master, Delta, ECS, ELAN, FocalTech, GIGABYTE, G.Skill, Innodisk, Intel, Inventec, InWin, ITRI, KIOXIA, Kneron, MediaTek, Microip, Mitac, MSI, NXP, NVIDIA, PEGATRON, Phison, PNY, PSMC, QCT, Qualcomm, Quanta, Realtek, RETRONIX, Seagate, Silicon Motion, Silicon Power, SOLOMON, Supermicro, SYSGRATION, TADA, Taiwan Micro, Thermaltake, Transcend, Wiwynn, and more. In total, 1500 local and international manufacturers will join COMPUTEX 2024 in not only the exhibition, but also forum and keynote speeches as well.

Event Dates: June 4 – 7, 2024
Event Location: Taipei Nangang Exhibition Center (TaiNEX) Halls 1 & 2
Main Theme: Connecting AI
Main Topics: AI computing, advanced communications, future mobility, immersive reality, innovation, green energy, and sustainability.

【Special Events】

• Pre-show Forum: May 16, 2024 [AI PC Industry Exploration Forum] Location: Grand HiLai Taipei Hotel
• Official Award Winners Announcement: May 28, 2024 [BC Awards 2024 Award Winning Products Announcement] Location: Le Méridien Taipei

COMPUTEX Official Website: https://www.computex.biz/
Event information and pre-registration website: https://show.computex.biz/
COMPUTEX CYBERWORLD website: https://show.computex.biz/online.aspx
Facebook: https://www.facebook.com/ComputexTaipei
YouTube Channel: https://www.youtube.com/user/COMPUTEXTAIPEIshow/
LinkedIn: https://www.linkedin.com/company/computex-taipei/